US8725506B2 - Speech audio processing - Google Patents

Speech audio processing Download PDF

Info

Publication number
US8725506B2
US8725506B2 US12/828,195 US82819510A US8725506B2 US 8725506 B2 US8725506 B2 US 8725506B2 US 82819510 A US82819510 A US 82819510A US 8725506 B2 US8725506 B2 US 8725506B2
Authority
US
United States
Prior art keywords
noise
speech
audio
speaker
variable
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related, expires
Application number
US12/828,195
Other languages
English (en)
Other versions
US20120004909A1 (en
Inventor
Willem M. Beltman
Matias Zanartu
Arijit Raychowdhury
Anand P. Rangarajan
Michael E. Deisher
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Intel Corp
Original Assignee
Intel Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to US12/828,195 priority Critical patent/US8725506B2/en
Application filed by Intel Corp filed Critical Intel Corp
Priority to EP11801384.6A priority patent/EP2589047A4/en
Priority to JP2013513424A priority patent/JP5644013B2/ja
Priority to PCT/US2011/042515 priority patent/WO2012003269A2/en
Priority to TW100123111A priority patent/TWI455112B/zh
Priority to KR1020127031843A priority patent/KR101434083B1/ko
Priority to CN201180027602.0A priority patent/CN102934159B/zh
Publication of US20120004909A1 publication Critical patent/US20120004909A1/en
Assigned to INTEL CORPORATION reassignment INTEL CORPORATION ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: BELTMAN, WILLEM M., DEISHER, MICHAEL E., RANGARAJAN, ANAND P., RAYCHOWDHURY, ARIJIT, ZANARTU, MATIAS
Application granted granted Critical
Publication of US8725506B2 publication Critical patent/US8725506B2/en
Expired - Fee Related legal-status Critical Current
Adjusted expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/28—Constructional details of speech recognition systems
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/20—Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0264—Noise filtering characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • G10L21/0224—Processing in the time domain
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • G10L21/0232—Processing in the frequency domain
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/93—Discriminating between voiced and unvoiced parts of speech signals

Definitions

  • the present invention relates generally to audio processing and in particular, to speech signal processing.
  • FIG. 1 is a diagram for a speech processing engine in accordance with some embodiments.
  • FIG. 2 is a diagram of a synthesizer in accordance with some embodiments.
  • FIG. 3 is a diagram of a structure for implementing a speech processing engine in accordance with some embodiments.
  • FIG. 4 is a diagram of an electronic device platform in accordance with some embodiments.
  • Voice command and continuous speech recognition are used for mobile Internet devices, for example, with in-car applications and phones that have limited keyboard functionality. It is desirable to be able to provide clean input to any speech recognition engine, but background noise in the environment impedes this objective. For example, experiments have shown that the open dictation word accuracy can degrade to approximately 20% in car noise and cafeteria environments, which may be unacceptable to the user.
  • Kalman filtering techniques may be used to improve speech signal processing.
  • speech recognition performance may be enhanced by bifurcating audio noise filtering processing into separate speech recognition and human reception paths. That is, the audio path may be cloned to generate a “perception” (or auditory reception) channel and a separate channel that is used for preprocessing audio for the speech recognition engine.
  • the audio path may be cloned to generate a “perception” (or auditory reception) channel and a separate channel that is used for preprocessing audio for the speech recognition engine.
  • FIG. 1 is a block diagram of a speech processing engine 102 in accordance with some embodiments. It comprises a Kalman based filtering engine 104 , a speaker/voice model 106 , an environmental noise model 107 , an automatic speech recognition (ASR) engine 108 , and a standard noise suppression block 110 .
  • ASR automatic speech recognition
  • Audio comes into the SPE (speech processing engine) and is split into two paths: a speech recognition path, entering the Kalman filter block 104 , and an audio perception path (cloned audio) that is processed using standard noise suppression techniques in block 110 for reception by a user.
  • the Kalman filter utilizes components from the speaker/voice model 106 , as well as from the environmental noise model 107 , to filter out noise from the audio signal and provide a filtered signal to the automatic speech recognition (ASR) engine 108 .
  • ASR automatic speech recognition
  • the speaker/noise model 106 (at least an initial version) is generated before SPE execution since the SPE works off of it, although an initial version may be fairly bare, and the speech/voice model may be updated while the SPE is executing.
  • the speaker/voice engine 106 provides particular characteristics associated with the current speaker. Such characteristics could include one or more glottal harmonics, including the user's particular fundamental glottal frequency, along with any other suitable information. For example, if previously acquired models (e.g., resulting from user training) are available, they may also be incorporated into the speaker/user model 106 . As indicated, previously generated “clean” audio information (x′(n)) for the particular user may also be used.
  • the environmental noise model 107 may be based on initial default data/assumptions for assumed noise environments or for specific or previously characterized environments (e.g., an office, a car, an airplane, etc). It may be static data (e.g., assumed background noise elements) associated with an environment and/or it may comprise dynamic data obtained from real-time sensors and the like. For example, it could include sensor inputs such as car speed, background noise microphone data, and air conditioning information, to enhance the performance of the noise model estimator. In some embodiments, a noise estimation method may be employed, e.g., for a single channel, by detecting periods of speech absence using a voice activity detector algorithm. The noise model may be further enhanced using an iterative loop between the noise model and Kalman filtering.
  • initial default data/assumptions for assumed noise environments or for specific or previously characterized environments (e.g., an office, a car, an airplane, etc). It may be static data (e.g., assumed background noise elements) associated with an environment and/or it may comprise
  • the filter 104 may use either or both the speaker model and noise model to filter the received audio signal. Again, from the speaker model, it may use an extension to add periodic components in the form of pulses into the Kalman filtering to account for glottal harmonics generated by the speech source (e.g., human or other entity speaker using, for example, a dictation, voice controlled, or translation device). Kalman filtering has typically been used with a white noise input, but in the case of human speech, the addition of a periodic input may more closely resemble the physiology of speech generation.
  • the speaker model information including the predetermined model information and glottal harmonic parameters may be used to load a set of predetermined or previously determined coefficients for the speaker model.
  • Kalman filtering results in audio that does not necessarily noticeably improve the human perception, but it does typically improve the performance of the speech recognition engine. Therefore, the audio path is cloned (two paths) to maximize both human perception and the speech recognition input using the Kalman pre processing filtering.
  • An implemented filter 104 using Kalman techniques can be used to model the vocal tract response as an AR or ARMA system, using an independent input and a driving noise, along with a noisy observation that accounts for additive colored-noise.
  • the driving periodic input is typically neglected and only a driving white noise is used for simplicity.
  • This assumption implies that the filter will (under an ideal performance) produce a clean but unvoiced speech signal, which neither has physiological value nor sounds natural.
  • the assumption may be adequate in cases where only filter parameters are needed.
  • the linear Kalman filter may capture the fundamental interactive features observed in voice production, thus yielding better estimates of the clean input under noisy conditions.
  • it may perform even better for speech processing applications.
  • the error in a scheme of this nature will be associated to its parameter estimation errors and not the product of a physiological/acoustical misrepresentation. Therefore, speech enhancement schemes disclosed herein are based on the linear Kalman filter, with the structure shown in the following table under the “Linear” heading.
  • the state x k corresponds to the clean speech input that is produced by the glottal source u k and environmental noise w k . (x, is not an actual input to the SPE.)
  • the measured signal y k is corrupted by the observation noise v k .
  • previous Kalman approaches neglect the periodic input u k for simplicity, yielding white noise excited speech.
  • the inclusion of such a periodic input and CP representation of the state transition matrix provides better estimates of the clean input x k and thus better speech recognition performance.
  • Kalman filtering as applied herein, will be discussed in more detail.
  • a Kalman filtering model-based approach is used for speech enhancement. It assumes that the clean speech follows a particular representation that is linearly corrupted with background noise. With standard Kalman filtering, clean speech is typically represented using an autoregressive (AR) model, which normally has a white Gaussian noise as an input. This is represented in discrete time equation 1.
  • AR autoregressive
  • x k+1 and x k are vectors containing p samples of the future and current clean speech
  • ⁇ is the state transition matrix that contains the LPC coefficients in the last row of a controllable canonical form
  • w k represents the white noise input that is converted into a vector that affects the current sample via the vector gain G.
  • the clean speech is projected via the projector vector H to obtain the current sample that is linearly added to the background noise v k to produce the corrupted observation or noisy speech y k .
  • Kalman filtering comprises two basic steps, a propagation step and an update step.
  • the model is used to predict the current sample based on the previous estimate (hence the notation n
  • the update step is depicted in equations (5)-(7), where the predicted samples are first corrected considering the error between the prediction and the estimate. This error is controlled by the Kalman gain K n , which is defined in equations (6) and (7). Note that all these parameters may be computed once within each frame, i.e., speech is considered a stationary process within each frame (normally of duration no longer than 25 ms).
  • the “modified Kalman filter” that is proposed in this project extends the standard filter by generalizing the two basic noise assumptions in the system, i.e., assuming that glottal pulses also drive the AR model during voiced segments and that the background noise has resonances associated with it (non-white process).
  • the glottal pulses are represented by u[n] and are present when there is vocal fold vibration.
  • the state equation needed to the Kalman filter can be extended by creating two subsystems embedded in a larger diagonal matrix.
  • the same system structure is used to track speech and noise as shown in equations (10) to (13), where the subscript s indicates speech and v indicates background noise.
  • the glottal pulses are introduced only in the current sample, for which the vector B has the same structure as G.
  • x k + 1 ⁇ ⁇ ⁇ x k + Bu k + Gw k ( 10 )
  • y k Hx k + v k ( 11 )
  • ⁇ [ ⁇ s 0 0 ⁇ v ] ( 12 )
  • H [ H s ⁇ ⁇ H v ] ( 13 )
  • the filter better represents speech signal and background noise conditions, thus yielding better noise removal and ASR performance.
  • the new Kalman filtering technique can not only be used for enhancement of speech recognition, but also to improve speech synthesis.
  • FIG. 2 a diagram showing a time-domain based synthesizer is shown.
  • the proposed scheme has a design that combines three interconnected processes that are applied to the input signal.
  • the first branch identifies the nature of the source component and creates a source signal.
  • the second branch searches for the filter structure and applies either CP-(closed phase) analysis or full-frame analysis to define the Linear Prediction Coefficients (LPC) of the filter.
  • LPC Linear Prediction Coefficients
  • the third branch detects the envelope and ensures stability of the synthetic sound.
  • branches can be computed in a sequential or parallel fashion and may use different frame and windowing structures (e.g., in some implementations, the first branch could use a rectangular window and non-overlapping frames, while the second one could use Hamming with, for example, 50% of overlap) as long the level of interaction is handled properly.
  • FIG. 3 shows a general structure for implementing a front-end for an audio processing engine, e.g., in a mobile device, for reducing power consumption. It illustrates a power efficient way to structure the different blocks, e.g., for the SPE 102 of FIG. 1 . It is divided into a compute intensive block 301 and a backend 305 , which is memory access intensive.
  • the compute intensive front end 301 has a filter processing section 302 and a decision block 304 for determining if input audio has speech within it.
  • the memory intensive back end 305 has speaker model block 306 for generating and updating the speaker model and a speech recognition block 308 for implementing ASR.
  • the speaker model block 306 may also have a noise model section for generating all or part of the noise model. Audio comes into the front end 301 , processed by filter 302 and if it has speech, as decided at decision block 304 , the speaker model and speech recognition blocks 306 , 308 are activated for processing the filtered speech signal from the filter 302 .
  • the hardware implementation of the speech enhancement algorithms in the front-end 301 provides opportunity for achieving low power and will also enables the use of a threshold detector 304 to provide a wake-up signal to the back-end of the processor hardware.
  • the back end 305 provides hardware implementation of the speech recognition algorithms e.g., (HMM and/or neural networks based), which is typically memory intensive, and high performance.
  • HMM and/or neural networks based the speech recognition algorithms
  • FIG. 4 shows an example of an electronic device platform 402 such as for a portable computing device, smart phone, and the like.
  • the represented portion comprises one or more processing cores 404 , graphics processor (GPX) 406 , memory controller hub (MCH) 408 , IO section 410 , and power management section 416 .
  • the GPX 406 interfaces with a display 407 to provide video content.
  • the MCH 408 interfaces with memory 409 for providing the platform with additional memory (e.g., volatile or non-volatile).
  • the power management section 416 controls a power source (e.g., battery, adapter converters, VRs, etc.) to provide power to the different platform sections, and it also manages the different activity states for reducing power consumption when feasible.
  • a power source e.g., battery, adapter converters, VRs, etc.
  • the IO section 410 comprises an audio processing section 412 and peripheral interfaces(s) 414 .
  • the Peripheral interface(s) provide interfaces (e.g., PCI, USB) for communicating and enabling various different peripheral devices 415 (keyboard, wireless interface, printer, etc.).
  • the audio processing section 412 may receive various audio input/output (analog and/or digital) for providing/receiving audio content from a user. It may also communicate with internal modules, for example, to communicate audio between a user and a network (e.g., cell, Internet, etc.).
  • the audio processing section 412 includes the various components (e.g., A/D/A converters, codecs, etc. for processing audio as dictated by the functions of the platform 402 .
  • the audio Px 412 includes an SPE 413 , as discussed herein, for implementing speech processing. In particular, it may comprise a power efficient structure as described in FIG. 3 .
  • Coupled is used to indicate that two or more elements are in direct physical or electrical contact with each other.
  • Connected is used to indicate that two or more elements are in direct physical or electrical contact with each other.
  • Connected is used to indicate that two or more elements are in direct physical or electrical contact with each other.
  • Connected is used to indicate that two or more elements are in direct physical or electrical contact with each other.
  • Coupled is used to indicate that two or more elements co-operate or interact with each other, but they may or may not be in direct physical or electrical contact.
  • PMOS transistor refers to a P-type metal oxide semiconductor field effect transistor.
  • NMOS transistor refers to an N-type metal oxide semiconductor field effect transistor.
  • MOS transistor MOS transistor
  • NMOS transistor NMOS transistor
  • PMOS transistor PMOS transistor
  • transistor can include other suitable transistor types, e.g., junction-field-effect transistors, bipolar-junction transistors, metal semiconductor FETs, and various types of three dimensional transistors, MOS or otherwise, known today or not yet developed.
  • suitable transistor types e.g., junction-field-effect transistors, bipolar-junction transistors, metal semiconductor FETs, and various types of three dimensional transistors, MOS or otherwise, known today or not yet developed.
  • IC semiconductor integrated circuit
  • PDA programmable logic arrays
  • signal conductor lines are represented with lines. Some may be thicker, to indicate more constituent signal paths, have a number label, to indicate a number of constituent signal paths, and/or have arrows at one or more ends, to indicate primary information flow direction. This, however, should not be construed in a limiting manner. Rather, such added detail may be used in connection with one or more exemplary embodiments to facilitate easier understanding of a circuit. Any represented signal lines, whether or not having additional information, may actually comprise one or more signals that may travel in multiple directions and may be implemented with any suitable type of signal scheme, e.g., digital or analog lines implemented with differential pairs, optical fiber lines, and/or single-ended lines.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Quality & Reliability (AREA)
  • Soundproofing, Sound Blocking, And Sound Damping (AREA)
  • Circuit For Audible Band Transducer (AREA)
  • Fittings On The Vehicle Exterior For Carrying Loads, And Devices For Holding Or Mounting Articles (AREA)
  • Circuits Of Receivers In General (AREA)
US12/828,195 2010-06-30 2010-06-30 Speech audio processing Expired - Fee Related US8725506B2 (en)

Priority Applications (7)

Application Number Priority Date Filing Date Title
US12/828,195 US8725506B2 (en) 2010-06-30 2010-06-30 Speech audio processing
JP2013513424A JP5644013B2 (ja) 2010-06-30 2011-06-30 発話音声処理
PCT/US2011/042515 WO2012003269A2 (en) 2010-06-30 2011-06-30 Speech audio processing
TW100123111A TWI455112B (zh) 2010-06-30 2011-06-30 語音處理設備及電子裝置
EP11801384.6A EP2589047A4 (en) 2010-06-30 2011-06-30 AUDIO PROCESSING OF THE WORD
KR1020127031843A KR101434083B1 (ko) 2010-06-30 2011-06-30 음성 오디오 처리
CN201180027602.0A CN102934159B (zh) 2010-06-30 2011-06-30 语音音频处理

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US12/828,195 US8725506B2 (en) 2010-06-30 2010-06-30 Speech audio processing

Publications (2)

Publication Number Publication Date
US20120004909A1 US20120004909A1 (en) 2012-01-05
US8725506B2 true US8725506B2 (en) 2014-05-13

Family

ID=45400342

Family Applications (1)

Application Number Title Priority Date Filing Date
US12/828,195 Expired - Fee Related US8725506B2 (en) 2010-06-30 2010-06-30 Speech audio processing

Country Status (7)

Country Link
US (1) US8725506B2 (ja)
EP (1) EP2589047A4 (ja)
JP (1) JP5644013B2 (ja)
KR (1) KR101434083B1 (ja)
CN (1) CN102934159B (ja)
TW (1) TWI455112B (ja)
WO (1) WO2012003269A2 (ja)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104463841A (zh) * 2014-10-21 2015-03-25 深圳大学 衰减系数自适应的滤波方法及滤波***

Families Citing this family (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8725506B2 (en) 2010-06-30 2014-05-13 Intel Corporation Speech audio processing
US8812014B2 (en) * 2010-08-30 2014-08-19 Qualcomm Incorporated Audio-based environment awareness
US9947333B1 (en) * 2012-02-10 2018-04-17 Amazon Technologies, Inc. Voice interaction architecture with intelligent background noise cancellation
US9437213B2 (en) 2012-03-05 2016-09-06 Malaspina Labs (Barbados) Inc. Voice signal enhancement
US9020818B2 (en) 2012-03-05 2015-04-28 Malaspina Labs (Barbados) Inc. Format based speech reconstruction from noisy signals
US9384759B2 (en) * 2012-03-05 2016-07-05 Malaspina Labs (Barbados) Inc. Voice activity detection and pitch estimation
US20140358552A1 (en) * 2013-05-31 2014-12-04 Cirrus Logic, Inc. Low-power voice gate for device wake-up
US9361890B2 (en) * 2013-09-20 2016-06-07 Lenovo (Singapore) Pte. Ltd. Context-based audio filter selection
US9413434B2 (en) 2013-10-04 2016-08-09 Intel Corporation Cancellation of interfering audio on a mobile device
KR101912177B1 (ko) 2013-11-15 2018-10-26 인텔 코포레이션 음성 인식 동적 사전을 유지하기 위한 시스템 및 방법
US9449602B2 (en) * 2013-12-03 2016-09-20 Google Inc. Dual uplink pre-processing paths for machine and human listening
KR102216048B1 (ko) 2014-05-20 2021-02-15 삼성전자주식회사 음성 명령 인식 장치 및 방법
US9858922B2 (en) 2014-06-23 2018-01-02 Google Inc. Caching speech recognition scores
US9299347B1 (en) 2014-10-22 2016-03-29 Google Inc. Speech recognition using associative mapping
US9786270B2 (en) 2015-07-09 2017-10-10 Google Inc. Generating acoustic models
US10229672B1 (en) 2015-12-31 2019-03-12 Google Llc Training acoustic models using connectionist temporal classification
EP3217399B1 (en) * 2016-03-11 2018-11-21 GN Hearing A/S Kalman filtering based speech enhancement using a codebook based approach
DE102017209585A1 (de) * 2016-06-08 2017-12-14 Ford Global Technologies, Llc System und verfahren zur selektiven verstärkung eines akustischen signals
US20180018973A1 (en) 2016-07-15 2018-01-18 Google Inc. Speaker verification
US10706840B2 (en) 2017-08-18 2020-07-07 Google Llc Encoder-decoder models for sequence to sequence mapping
WO2019169616A1 (zh) * 2018-03-09 2019-09-12 深圳市汇顶科技股份有限公司 语音信号处理方法及装置
CN110738990B (zh) * 2018-07-19 2022-03-25 南京地平线机器人技术有限公司 识别语音的方法和装置
WO2021041741A1 (en) * 2019-08-30 2021-03-04 Dolby Laboratories Licensing Corporation Pre-conditioning audio for machine perception
GB202104280D0 (en) * 2021-03-26 2021-05-12 Samsung Electronics Co Ltd Method and apparatus for real-time sound enhancement
CN113053382B (zh) * 2021-03-30 2024-06-18 联想(北京)有限公司 处理方法和装置

Citations (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5148488A (en) * 1989-11-17 1992-09-15 Nynex Corporation Method and filter for enhancing a noisy speech signal
US5570453A (en) * 1993-02-23 1996-10-29 Motorola, Inc. Method for generating a spectral noise weighting filter for use in a speech coder
TW309675B (en) 1996-12-26 1997-07-01 Yiing Lii Method and apparatus for complex fuzzy signal processing
US5774846A (en) * 1994-12-19 1998-06-30 Matsushita Electric Industrial Co., Ltd. Speech coding apparatus, linear prediction coefficient analyzing apparatus and noise reducing apparatus
US5864810A (en) * 1995-01-20 1999-01-26 Sri International Method and apparatus for speech recognition adapted to an individual speaker
US5915234A (en) * 1995-08-23 1999-06-22 Oki Electric Industry Co., Ltd. Method and apparatus for CELP coding an audio signal while distinguishing speech periods and non-speech periods
US5970446A (en) * 1997-11-25 1999-10-19 At&T Corp Selective noise/channel/coding models and recognizers for automatic speech recognition
TW425542B (en) 1999-03-19 2001-03-11 Ind Tech Res Inst Kalman filter for speech enhancement
JP2002006898A (ja) 2000-06-22 2002-01-11 Asahi Kasei Corp ノイズ低減方法及びノイズ低減装置
US20020026253A1 (en) 2000-06-02 2002-02-28 Rajan Jebu Jacob Speech processing apparatus
US20020055913A1 (en) 2000-06-02 2002-05-09 Rajan Jebu Jacob Signal processing system
US6408269B1 (en) * 1999-03-03 2002-06-18 Industrial Technology Research Institute Frame-based subband Kalman filtering method and apparatus for speech enhancement
US6427134B1 (en) * 1996-07-03 2002-07-30 British Telecommunications Public Limited Company Voice activity detector for calculating spectral irregularity measure on the basis of spectral difference measurements
US20020120443A1 (en) * 2001-02-28 2002-08-29 Ibm Corporation Speech recognition in noisy environments
US20030177007A1 (en) * 2002-03-15 2003-09-18 Kabushiki Kaisha Toshiba Noise suppression apparatus and method for speech recognition, and speech recognition apparatus and method
US20040064315A1 (en) * 2002-09-30 2004-04-01 Deisher Michael E. Acoustic confidence driven front-end preprocessing for speech recognition in adverse environments
US20040083100A1 (en) * 1996-02-06 2004-04-29 The Regents Of The University Of California System and method for characterizing voiced excitations of speech and acoustic signals, removing acoustic noise from speech, and synthesizing speech
US6757651B2 (en) * 2001-08-28 2004-06-29 Intellisist, Llc Speech detection system and method
US20040193411A1 (en) * 2001-09-12 2004-09-30 Hui Siew Kok System and apparatus for speech communication and speech recognition
KR20050106235A (ko) 2004-05-04 2005-11-09 주식회사 팬택앤큐리텔 단말기에서의 에코 및 잡음 제거 장치
US7072833B2 (en) * 2000-06-02 2006-07-04 Canon Kabushiki Kaisha Speech processing system
US7082393B2 (en) * 2001-03-27 2006-07-25 Rast Associates, Llc Head-worn, trimodal device to increase transcription accuracy in a voice recognition system and to process unvocalized speech
US7117157B1 (en) * 1999-03-26 2006-10-03 Canon Kabushiki Kaisha Processing apparatus for determining which person in a group is speaking
US20080288258A1 (en) 2007-04-04 2008-11-20 International Business Machines Corporation Method and apparatus for speech analysis and synthesis
US7457750B2 (en) * 2000-10-13 2008-11-25 At&T Corp. Systems and methods for dynamic re-configurable speech recognition
US20090076814A1 (en) * 2007-09-19 2009-03-19 Electronics And Telecommunications Research Institute Apparatus and method for determining speech signal
US20090163168A1 (en) * 2005-04-26 2009-06-25 Aalborg Universitet Efficient initialization of iterative parameter estimation
US20090222263A1 (en) * 2005-06-20 2009-09-03 Ivano Salvatore Collotta Method and Apparatus for Transmitting Speech Data To a Remote Device In a Distributed Speech Recognition System
WO2009116291A1 (ja) 2008-03-21 2009-09-24 学校法人東京理科大学 雑音抑圧装置および雑音抑圧方法
US20100131269A1 (en) * 2008-11-24 2010-05-27 Qualcomm Incorporated Systems, methods, apparatus, and computer program products for enhanced active noise cancellation
US20110054892A1 (en) * 2008-05-28 2011-03-03 Koreapowervoice Co., Ltd. System for detecting speech interval and recognizing continuous speech in a noisy environment through real-time recognition of call commands
US20110077939A1 (en) * 2009-09-30 2011-03-31 Electronics And Telecommunications Research Institute Model-based distortion compensating noise reduction apparatus and method for speech recognition
US20110125490A1 (en) * 2008-10-24 2011-05-26 Satoru Furuta Noise suppressor and voice decoder
US20110305345A1 (en) * 2009-02-03 2011-12-15 University Of Ottawa Method and system for a multi-microphone noise reduction
WO2012003269A2 (en) 2010-06-30 2012-01-05 Intel Corporation Speech audio processing
US8121837B2 (en) * 2008-04-24 2012-02-21 Nuance Communications, Inc. Adjusting a speech engine for a mobile computing device based on background noise
US20120179462A1 (en) * 2007-07-06 2012-07-12 David Klein System and Method for Adaptive Intelligent Noise Suppression

Patent Citations (39)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5148488A (en) * 1989-11-17 1992-09-15 Nynex Corporation Method and filter for enhancing a noisy speech signal
US5570453A (en) * 1993-02-23 1996-10-29 Motorola, Inc. Method for generating a spectral noise weighting filter for use in a speech coder
US5774846A (en) * 1994-12-19 1998-06-30 Matsushita Electric Industrial Co., Ltd. Speech coding apparatus, linear prediction coefficient analyzing apparatus and noise reducing apparatus
US6205421B1 (en) * 1994-12-19 2001-03-20 Matsushita Electric Industrial Co., Ltd. Speech coding apparatus, linear prediction coefficient analyzing apparatus and noise reducing apparatus
US5864810A (en) * 1995-01-20 1999-01-26 Sri International Method and apparatus for speech recognition adapted to an individual speaker
US5915234A (en) * 1995-08-23 1999-06-22 Oki Electric Industry Co., Ltd. Method and apparatus for CELP coding an audio signal while distinguishing speech periods and non-speech periods
US20040083100A1 (en) * 1996-02-06 2004-04-29 The Regents Of The University Of California System and method for characterizing voiced excitations of speech and acoustic signals, removing acoustic noise from speech, and synthesizing speech
US6427134B1 (en) * 1996-07-03 2002-07-30 British Telecommunications Public Limited Company Voice activity detector for calculating spectral irregularity measure on the basis of spectral difference measurements
TW309675B (en) 1996-12-26 1997-07-01 Yiing Lii Method and apparatus for complex fuzzy signal processing
US5970446A (en) * 1997-11-25 1999-10-19 At&T Corp Selective noise/channel/coding models and recognizers for automatic speech recognition
US6408269B1 (en) * 1999-03-03 2002-06-18 Industrial Technology Research Institute Frame-based subband Kalman filtering method and apparatus for speech enhancement
TW425542B (en) 1999-03-19 2001-03-11 Ind Tech Res Inst Kalman filter for speech enhancement
US7117157B1 (en) * 1999-03-26 2006-10-03 Canon Kabushiki Kaisha Processing apparatus for determining which person in a group is speaking
US20020026253A1 (en) 2000-06-02 2002-02-28 Rajan Jebu Jacob Speech processing apparatus
US20020055913A1 (en) 2000-06-02 2002-05-09 Rajan Jebu Jacob Signal processing system
US7072833B2 (en) * 2000-06-02 2006-07-04 Canon Kabushiki Kaisha Speech processing system
JP2002006898A (ja) 2000-06-22 2002-01-11 Asahi Kasei Corp ノイズ低減方法及びノイズ低減装置
US7457750B2 (en) * 2000-10-13 2008-11-25 At&T Corp. Systems and methods for dynamic re-configurable speech recognition
US20020120443A1 (en) * 2001-02-28 2002-08-29 Ibm Corporation Speech recognition in noisy environments
US7082393B2 (en) * 2001-03-27 2006-07-25 Rast Associates, Llc Head-worn, trimodal device to increase transcription accuracy in a voice recognition system and to process unvocalized speech
US6757651B2 (en) * 2001-08-28 2004-06-29 Intellisist, Llc Speech detection system and method
US20040193411A1 (en) * 2001-09-12 2004-09-30 Hui Siew Kok System and apparatus for speech communication and speech recognition
US20030177007A1 (en) * 2002-03-15 2003-09-18 Kabushiki Kaisha Toshiba Noise suppression apparatus and method for speech recognition, and speech recognition apparatus and method
US20040064315A1 (en) * 2002-09-30 2004-04-01 Deisher Michael E. Acoustic confidence driven front-end preprocessing for speech recognition in adverse environments
KR20050106235A (ko) 2004-05-04 2005-11-09 주식회사 팬택앤큐리텔 단말기에서의 에코 및 잡음 제거 장치
US20090163168A1 (en) * 2005-04-26 2009-06-25 Aalborg Universitet Efficient initialization of iterative parameter estimation
US20090222263A1 (en) * 2005-06-20 2009-09-03 Ivano Salvatore Collotta Method and Apparatus for Transmitting Speech Data To a Remote Device In a Distributed Speech Recognition System
US20080288258A1 (en) 2007-04-04 2008-11-20 International Business Machines Corporation Method and apparatus for speech analysis and synthesis
US20120179462A1 (en) * 2007-07-06 2012-07-12 David Klein System and Method for Adaptive Intelligent Noise Suppression
US20090076814A1 (en) * 2007-09-19 2009-03-19 Electronics And Telecommunications Research Institute Apparatus and method for determining speech signal
WO2009116291A1 (ja) 2008-03-21 2009-09-24 学校法人東京理科大学 雑音抑圧装置および雑音抑圧方法
US20100262425A1 (en) * 2008-03-21 2010-10-14 Tokyo University Of Science Educational Foundation Administrative Organization Noise suppression device and noise suppression method
US8121837B2 (en) * 2008-04-24 2012-02-21 Nuance Communications, Inc. Adjusting a speech engine for a mobile computing device based on background noise
US20110054892A1 (en) * 2008-05-28 2011-03-03 Koreapowervoice Co., Ltd. System for detecting speech interval and recognizing continuous speech in a noisy environment through real-time recognition of call commands
US20110125490A1 (en) * 2008-10-24 2011-05-26 Satoru Furuta Noise suppressor and voice decoder
US20100131269A1 (en) * 2008-11-24 2010-05-27 Qualcomm Incorporated Systems, methods, apparatus, and computer program products for enhanced active noise cancellation
US20110305345A1 (en) * 2009-02-03 2011-12-15 University Of Ottawa Method and system for a multi-microphone noise reduction
US20110077939A1 (en) * 2009-09-30 2011-03-31 Electronics And Telecommunications Research Institute Model-based distortion compensating noise reduction apparatus and method for speech recognition
WO2012003269A2 (en) 2010-06-30 2012-01-05 Intel Corporation Speech audio processing

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
International Preliminary Report on Patentability received for PCT application No. PCT/US2011/042515, mailed on Jan. 17, 2013, 5 pages.
International Search Report and Written Opinon received for PCT application No. PCT/US2011/042515, mailed on Feb. 9, 2012, 8 Pages.
Notice of Preliminary Rejection for Korean Patent Application No. 10-2012-7031843 received Feb. 7, 2014, 7 pages including 3 pages of English translation.
Office Action for Taiwanese Patent Application 100123111, mailed on Sep. 26, 2013. 11 pages including 7 pages of English language translation.
Office Action Received for Japanese Patent Application No. 2013-513424, mailed on Dec. 10, 2013, 4 pages of Office Action including 2 pages of English Translation.
Thomas F. Quatieri, Discrete-time Speech Signal Processing Principle and Practice, Prentice Hall Signal Processing Series, 2002; pp. 59, 64, 225-226, 150. *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104463841A (zh) * 2014-10-21 2015-03-25 深圳大学 衰减系数自适应的滤波方法及滤波***

Also Published As

Publication number Publication date
TW201222527A (en) 2012-06-01
EP2589047A2 (en) 2013-05-08
WO2012003269A3 (en) 2012-03-29
JP2013531275A (ja) 2013-08-01
KR20130033372A (ko) 2013-04-03
US20120004909A1 (en) 2012-01-05
WO2012003269A2 (en) 2012-01-05
KR101434083B1 (ko) 2014-08-25
EP2589047A4 (en) 2015-11-25
CN102934159B (zh) 2015-12-16
CN102934159A (zh) 2013-02-13
TWI455112B (zh) 2014-10-01
JP5644013B2 (ja) 2014-12-24

Similar Documents

Publication Publication Date Title
US8725506B2 (en) Speech audio processing
KR101224755B1 (ko) 음성-상태 모델을 사용하는 다중-감각 음성 향상
US8249270B2 (en) Sound signal correcting method, sound signal correcting apparatus and computer program
KR101004495B1 (ko) 증분 베이즈 학습을 사용하는 잡음 추정 방법
US9536538B2 (en) Method and device for reconstructing a target signal from a noisy input signal
KR20160125984A (ko) 화자 사전 기반 스피치 모델링을 위한 시스템들 및 방법들
US9378755B2 (en) Detecting a user's voice activity using dynamic probabilistic models of speech features
Cohen Speech enhancement using super-Gaussian speech models and noncausal a priori SNR estimation
US11308946B2 (en) Methods and apparatus for ASR with embedded noise reduction
Stouten et al. Model-based feature enhancement with uncertainty decoding for noise robust ASR
Wang et al. Mask estimation incorporating phase-sensitive information for speech enhancement
Ma et al. Combining speech fragment decoding and adaptive noise floor modeling
Vuppala et al. Recognition of consonant-vowel (CV) units under background noise using combined temporal and spectral preprocessing
Li et al. Robust log-energy estimation and its dynamic change enhancement for in-car speech recognition
Yoon et al. Speech enhancement based on speech/noise-dominant decision
Oonishi et al. A noise-robust speech recognition approach incorporating normalized speech/non-speech likelihood into hypothesis scores
Chen et al. Research on Speech Recognition of Sanitized Robot Based on Improved Speech Enhancement Algorithm
Sarafnia et al. Implementation of Bayesian recursive state-space Kalman filter for noise reduction of speech signal
Park et al. Noise reduction scheme for speech recognition in mobile devices
Setiawan Exploration and optimization of noise reduction algorithms for speech recognition in embedded devices
Singh et al. Speech Enhancement using Segmental Non-Negative Matrix Factorization (SNMF) and Hidden Marvok Model (HMM)
Thatphithakkul et al. Combined simulated data adaptation and piecewise linear transformation for robust speech recognition
Kleinschmidt et al. A likelihood-maximizing framework for enhanced in-car speech recognition based on speech dialog system interaction
JP2013068801A (ja) 音響モデル作成装置、音響モデル作成方法、及び音響モデル作成システム

Legal Events

Date Code Title Description
AS Assignment

Owner name: INTEL CORPORATION, CALIFORNIA

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BELTMAN, WILLEM M.;ZANARTU, MATIAS;RAYCHOWDHURY, ARIJIT;AND OTHERS;REEL/FRAME:027953/0105

Effective date: 20100908

FEPP Fee payment procedure

Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

STCF Information on status: patent grant

Free format text: PATENTED CASE

CC Certificate of correction
MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551)

Year of fee payment: 4

FEPP Fee payment procedure

Free format text: MAINTENANCE FEE REMINDER MAILED (ORIGINAL EVENT CODE: REM.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

LAPS Lapse for failure to pay maintenance fees

Free format text: PATENT EXPIRED FOR FAILURE TO PAY MAINTENANCE FEES (ORIGINAL EVENT CODE: EXP.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

STCH Information on status: patent discontinuation

Free format text: PATENT EXPIRED DUE TO NONPAYMENT OF MAINTENANCE FEES UNDER 37 CFR 1.362

FP Lapsed due to failure to pay maintenance fee

Effective date: 20220513